Thorough review by the International Agency for Research on Cancer (IARC) has resulted in classifying many substances, including pharmaceuticals, as probably or possibly carcinogenic to humans, based on experiments on animals or limited data on humans. We evaluated 9 such pharmaceuticals for evidence of carcinogenicity in patients receiving them in a large medical care program with automated pharmacy records and a cancer registry. Nested case-control analyses were performed in a cohort of 6.5 million subscribers with up to 12 years of follow-up, focusing on cancer sites suggested by previous evidence and other sites with odds ratio of at least 1.50, p < 0.01 and some evidence of dose-response. Unmeasured confounding was estimated in sensitivity analyses. We found some supportive evidence for carcinogenicity of griseofulvin, metronidazole and phenytoin and for the known carcinogen, cyclophosphamide, which was added for validation of our data and analyses. Findings for chloramphenicol, iron-dextran complex, phenoxybenzamine and phenobarbital were essentially non-contributory. Confounding by cigarette smoking and prior thyroid disease could account, respectively, for associations of oxazepam with lung cancer and propylthiouracil with thyroid cancer. Although not definitive, these findings should be considered in the evaluations of these pharmaceuticals.